Akademik Veri Yönetim Sistemi
BIOMASS & BIOENERGY, cilt.212, 2026 (SCI-Expanded, Scopus)
Switchgrass (Panicum virgatum L.) is a promising bioenergy crop due to its high biomass yield and broad adaptation to adverse soil and growing conditions. Accurate and rapid ploidy level determination is crucial for a successful switchgrass breeding program to produce new varieties with better quantitative and qualitative biomass traits, higher bioenergy production potential and broad adaptation to adverse growing conditions. Classical methods to assess the ploidy level and the DNA content of switchgrass rely on two main techniques: flow cytometry and cytogenetic chromosome counting but they necessitate advanced and expensive instruments and chemicals, complex protocols, intensive labor, and experienced staff, limiting their use in high-throughput breeding applications. In the present study, the effectiveness of Fourier-transform near-infrared spectroscopy (FT-NIRS) data combined with various data preprocessing treatments, feature selection methods and machine learning (ML) algorithms was tested for the first time to classify switchgrass samples (n = 239) into two classes according to their ploidy levels as tetraploid and octoploid. The best classification results were obtained from the combination of feature selection by Random Forest (RF) algorithm with the data preprocessed by using Z-score standardization and classification with the Support Vector Machines (SVM) procedure yielding a classification test accuracy of 87%. The study results indicate that the NIRS and ML algorithms elucidate promising results to rapidly and efficiently classify switchgrass plants based on their ploidy levels without the classical flow cytometry and cytogenetic chromosome counting methods.